1 / 29

Acknowledgements Methodology : Mike Charles, Zoltan Toth, Roman Krzysztofowicz

Operational Implementation of Climatologically Calibrated Precipitation Analysis (CCPA) ---- Statistical Calibration of STAGE IV Precipitation Data Set Dingchen Hou and Yuejian Zhu. Acknowledgements Methodology : Mike Charles, Zoltan Toth, Roman Krzysztofowicz

magee
Télécharger la présentation

Acknowledgements Methodology : Mike Charles, Zoltan Toth, Roman Krzysztofowicz

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Operational Implementation ofClimatologically Calibrated Precipitation Analysis (CCPA)----Statistical Calibration of STAGE IV Precipitation Data Set Dingchen Hou and Yuejian Zhu Acknowledgements Methodology: Mike Charles, Zoltan Toth, Roman Krzysztofowicz Software Contributions: Mike Charles, Malaquias Pena Data Sources: Ying Lin, Pingping Xie, Mike Charles, Letitia Soulliard Operational Requirements: Dong-Jun Seo, Zoltan Toth, Steve Lord

  2. What is Stage IV • National Stage IV QPE product • Mosaicked from Individual RFC's Multi-sensor Precipitation Analyses (RMPAs) • Available within 1h of receiving any new hourly/6-hourly data from one or more RFCs. • 12 RFCs over CONUS • Some manual QC at RFCs • 4km HRAP grid

  3. RFC precipitation data (mm, 6h acc.), Two Versions Stage IV Operational Stage IV Operational HPC Experimental HPC Experimental

  4. Stage IV Operational Stage IV Operational RFC precipitation data (mm, 6h acc.) by EMC/HPC HPC Experimental HPC Experimental

  5. Why Calibration ?(following Charles et. al) • Want an accurate, 5x5km (NDFD) 6-hourly precipitation for • Down scaling NAEFS precip forecast to NDFD • Verifying NAEFS precipitation forecast • Stage IV, a good candidate but … • High resolution (close to NDFD) → better representation of fine scale temporal and spatial variability • Non-uniform QC (different RFCs have different methods) • Each RFC may make their own adjustments before mosaicking • CPC Unified Precipitation Analysis • Back to 2000 (eventually back to 1979, then 1948) • ⅛° spatial resolution • Daily • Global land • More confidence in long term statistics of CPC dataset • Uniform QC across entire domain • Gauge-based • Too low resolution for downscaling

  6. Comparison of CPC and Stage IV CPC: CONUS LAND ONLY CPC: After Format Change Stage IV: HPC Exp. CONUS LAND Stage IV: Geometric Boundary

  7. How to Calibrate?(Adopted from Charles et. al) • Solution: adjust RFC grids so their climatology is consistent with the CPC dataset • Have the reliability of the CPC dataset, with the high spatial and temporal resolution of the RFC dataset • Note: This effort has limitations, as it was developed to simply combine existing datasets. Much more work will be needed for a more comprehensive approach, but this is out of the scope of this work

  8. Flow Chart 1/8 Degree ~5km STAGE IV At 1/8d (7 year historic) Aggregation to 1/8 degree resolution 24 hourly Adjustment (How to?) CPC_1/8d (7 year Archive) Down scaling Spatial disaggregation Accumulate To 24 hours Temporal Disaggregation 6 hourly Adjusted STAGE IV (output) STAGE IV (Input) STAGE IV (7 year Archive)

  9. Establish Statistical Relationship(Charles et. al) Historical data sets June 1. 2002 to July 31 For CPC and STAGE IV Match resolutions Accumulate RFC over 24 hours Interpolate to ⅛° (copygb w/ volume preservation) Collect precip samples For each day of the year and at each grid point, collect all precip within 60 day window centered around that day, over all 7 years (max ~427 data points) Use only data points with RFC > 0 Linear regression CPC=a⋅RFC+b End Result Linear relationship (a & b) on ⅛° grid for each day of the year 9

  10. Regression Aug. 1st(SW US, Summer Gaps, maximum 369 points) N a e b

  11. Time Series of regression(SW US, Summer Gaps) N a 1 100 e b 50 0

  12. Filling the gap in Space (linear interpolation)

  13. Time Series; Filing the Gap in Space a a 1 1 b b 0 0

  14. Regression Adjustment, raw and filled a&b

  15. Adjustment with raw and filled a&b

  16. Temporal Smoothing (3 harmonics) of a Neighboring Land point A costal grid point 100 1 Typical point Typical point 1 1

  17. Temporal Smoothing (3 harmonics) of b A costal grid point Neighboring Land point Typical point Typical point

  18. Recovering Original RFC Resolutions Information is lost between H and L How much information? Take ratio H/L (below) This ratio can be used to put high resolution information back into RFC* Interpolate CPC* to NDFD Multiply by H/L End with CPC* at NDFD resolution. Spatial information recovered from RFCorig H L Ratio (H/L) NDFD NDFD->⅛°->NDFD Spatial Disaggregation (Charles et. al) Daily RFC precip 18

  19. Recovering Original RFC Resolutions Daily Total: 4” Daily Total: 6” Temporal Disaggregation (Charles et. al) • Determine percentage of daily total precipitation in each 6-hour period in RFCorig • Divide 24 hour RFC* into four 6-hour precip amounts using the percentages from RFCorig Percent of daily total in each 6-hourly period RFCorig RFCadj 19

  20. Implementation Details Rules Only Non-Zero Stage IV is adjusted Zero values remains zero Adjustment is applied over CONUS LAND only Leap Year 366 day convention is adapted in regression calculations Feb 29 has its own regression coefficients a and b Spatial Continuity US Boundaries Land/Ocean Boundary Zero/Non-Zero Boundary Rare cases of abnormal regression coefficients Temporal smoothing of a and b reduces abnormal values Discard the regression coefficients a and b, if too large Set an upper limit to the adjusted St4 value

  21. What is next? • Operational implementation at NCEP, planned for 2010 • Generate the historical data set of CCPA for 2002-2010 • Real time generation of CCPA after STAGE IV, once per day • Periodic (annual) upgrading regression coefficients with increasing sample size • Updating coefficients a and b for real time CCPA • Re-generate the CCPA historical data set ? • Improving the methodology

  22. Quasi-Real Time DemonstrationsChristmas Day 2009

  23. STAGE IV 12/24 CCPA 12/24

  24. STAGE IV 12/25 CCPA 12/25

  25. STAGE IV 12/26 CCPA 12/26

  26. Quasi-Real Time DemonstrationSnowstorm 2010

  27. STAGE IV 02/04 CCPA 02/04

  28. STAGE IV 02/05 CCPA 02/05

  29. STAGE IV 02/06 CCPA 02/06

More Related